Cognitive Computing for Tacit Knowledge1
- 1. Cognitive Computing for Tacit Knowledge - Palliative or Tonic? Cognitive Computing Enthusiasts September 16th, 2015 HackerDojo, Mountain View, CA
- 2. About me
- 3. Cognitive Sciences: 1950 - 1978 Interdisciplinary inquiry of several university researchers/experts A report accepted by Sloan Foundation in 1978 (picture). Aimed at “universal science” to discover the representational and computational capacities of the human mind and their structural and functional realization in the human brain. The cognitive revolution: a historical perspective by George A. Miller, Department of Psychology, Princeton University (2003) In 1978: - cybernetics used concepts developed by computer science to model brain functions elucidated in neuroscience. - computer science and linguistics were linked through computational linguistics 1978 - 2004 Is cracking “human mind computation” the solution to semiconductor technology crisis? Why were “cognitive sciences” silent during the boom of personal computers, Internet and Enterprise IT? I am here to share & learn… Which “Wave” are we on now?
- 4. Cognition (Definition) “Set of all mental abilities and processes related to knowledge, attention, memory and working memory, judgment and evaluation, reasoning and "computation", problem solving and decision making, comprehension and production of language, etc…. Cognitive processes use existing knowledge and generate new knowledge” (Wikipedia, retrieved September 2015) Cognitive Computing (Adaptation) “Makes a new class of problems computable. It addresses complex situations that are characterized by ambiguity and uncertainty; in other words it handles human kinds of problems” (Wikipedia, retrieved September 2015) Cognitive Computing Systems (Purpose) “Learn and interact naturally with people to extend what either humans or machine could do on their own. They help human experts make better decisions by penetrating the complexity of Big Data” (IBM Research , September 2015) Identity? Humanities? “Socio-technical” systems?
- 5. Traditions of Cognitive Science COGNITIVISM EMERGENCE ENACTION "When the symbols appropriately represent some aspect of the real world, and the information processing leads to ... successful solution of the problem given..." "When the emergent properties (and resulting structure) can be seen to correspond to a specific cognitive capacity -- a successful solution to a required task." "When it becomes part of an ongoing existing world (as the young of every species do) or shapes a new one (as happens in evolutionary history)." Varela, Thompson & Rosch (The Embodied Mind : Cognitive Science and Human Experience, Cambridge, MA: MIT Press, 1991.1991). Encyclopaedia Autopoietica: a work of Randall Whitaker http://www.informatik.umu.se/~rwhit/EAIntro.html , retrieved from http://www.cybsoc.org/EA.html#enactive%20cognitive%20science on September 15th 2015 People learning through problem solving (with or without machines) People solving problems with machines (higher complexity, less self-learning, more stress) HOW DO I KNOW WHEN A COGNITIVE SYSTEM IS FUNCTIONING ADEQUATELY? (beyond the Turing test/Jeopardy game) A “Personal Watson” or a “Synthetic Neocortex”?
- 6. 2 mins window 200 bits storage & manipulation False memory syndrome Memory enrichment “Flash-bulb” memory Hyper-diffused storage We have in the cranium a slightly alkaline three-pound electrochemical computer running on glucose at about 25 watts. This computer contains some ten thousand million (that’s ten to the ten) logical elements called neurons, operating on a basic scanning rhythm of ten cycles per second. Then this is a high-variety dynamic system all right; but it really is finite. It follows (from Ashby’s Law) that we can recognize patterns up to a certain limit, and not beyond (Stafford Beer, “Science in the Service of Man” , 1973) Our System 2 requires at least 1.5 minutes (90 seconds) to engage Our System 2 has a very small size working memory: - most of us can recall only 5 to 9 of the items shown once Our System 2 has a “faulty” long term memory: - can not remember each and every time we need to remember - each time a memory is accessed, it is rebuilt and distorted I am intrigued that Cognitive Computing field does so little to help us deal with our weaknesses and so much to “mimic the way the human brain works” Conceptual differences ….
- 7. How CC helps me and you create value and preserve identity? Real-Time collaboration in creative work (PhD research& dissertation – 1994 -1999) Industrialization of Software Engineering, Internet and Enterprise IT (1996 onwards) Managing variability to develop healthy and productive organizations (2008 onwards) Clark, D.R. (2004). Knowledge Typology Map Retrieved from http://www.nwlink.com/~donclark/knowledge/knowledge_typology.html on June 2014. Knowledge is information that changes something or somebody— either by becoming grounds for actions, or by making an individual (or an institution) capable of different or more effective action." - Peter F. Drucker in The New Realities Tacit knowledge is personal knowledge embedded in individual experience and involves intangible factors, such as personal beliefs, perspective, and the value system. - Know-how and other cognitive sciences aspects - Hard to articulate with languages Environment of digital computers in all individual and institutional aspects
- 8. Cognitive computing for collaborative creative work in telepresence (analog or digital – does it matter?) Topic: "Group communication in a computer mediated environment: analysis, experiences and evaluations on enterprise collaborative projects and group training" Digital technologies context: Telecom ISDN (2*64kb/16kb, PictureTel/H323), CSCW (server centric design), OO Programming (principles), Internet (IETF), Mbone (ALF/ILP, VBR encoding, p2p principles, open source) Idea: “let each member of the group instantiate his/her own workspace, as he/she feels the need while engaged in solving the task at hand” (CS/Software Engineering R&D 1994-1997/PhD dissertation March 1999)
- 9. 9 Proposed model: “computer mediated telepresence” as a cognitive system - to the benefit of group accomplishment - rather than a multimodal communication system Inspired by Roberto Maturana’s Cognition (1975) how cognition as a biological phenomenon takes place and Maturana & Varela “autopoietic systems” “Can the whole system - people, computers, software tools - operate effectively and successfully in a given domain, language included?" Test and validation for specific domain (professional training) different group interactions to accomplish specific goals in situation of telepresence • Objective assessments (work results) • Compared with identical scenarios in physical presence • Subjective assessment (technology sufficiency/affordability/usability) Prototype & experimentation of above model for mediated cooperative work on practical case studies: Peer-to-peer architectures, User agents, Application Level Framing and Integrated Layer Processing for adaptation to variable conditions, Internet Mbone (IP multicast source based routing), variable bit-rate video encoding, (IETF MMUSIC + …. SIP adoption in 3GPP) NRG@ Cross the borders into cognitive sciences: It takes a village: 1995-1996 school year experimentation: 1 professor of Mechanical Engineering @INSA de Lyon, 40 students in 4 mediated sessions and 30 in 3 real presence sessions (1 semester of studies in “Genie Productique”/mechanical engineering ) + 1 professor of social sciences 3 scenarios: lecture, lab and project (teams of 5 people, isolated in media offices) Mbone
- 10. Lecture set up – very much like today’s webinars Memory test Concept test
- 11. 11 Qualité d’usage/rôle Apprenant Formateur 1 3,85 3,66 2 2,87 4,66 3 3,43 4 4a 4b 2,85 3,58 5 3 Qualité d’usage/rôle exécutant chef apprenant (exéc.+chef) Formateur 1 4,11 3,83 4,04 4,25 2 3,52 3,52 3,52 3,6 3 4,2 4,2 4,2 4,8 4a 4b 3,2 2,7 3,5 2,7 3,4 2,7 3,6 situation de cours/travaux dirigés situation de projet coopératif 1. l’acceptation 2. la difficulté d’exécution de la tâche 3. l’effort personnel 4.la qualité de réponse du système: 4a. pour la production 4b pour la communication/téléprésence Landed 3 years contract 1996-1999 in TeleRegions SUN – TeleApplications for Europeans Regions (health, education, citizen services in 6 regions, 4 countries) Project results (avg) Usability results (avg)
- 12. Biology and Human Behavior: The Neurological Origins of Individuality Robert Sapolsky, Ph.D. Professor of Neurology and Neurosurgery Stanford University 1996 2015 (still no video for everyone!) This experiment 20 years after …. But mobile operators are launching Video over LTE using …. IP Multicast, RTP/RTCP! It took 20 years of denial. Economists expertise required!
- 13. 1. How much of our creative work, the one that engages deeply our tacit knowledge, is “real-time collaborative”? - Writing, Coding or Designing/Planning/Creating/Simulating an action or an object - Manufacturing/Acting on physical objects - Pondering alternatives and deciding 2. Can we work & communicate synchronously such as our communicative behavior becomes a mutual orientation for the purpose of accomplishing a creative task rather than a painful exercise of memory and manipulation of widgets? We have: • Online gaming for skills development (procedural knowledge) • Webex/Citrix/Skype/Whatsapp/WeChat/IM for group communication (presence and telecommunications) • Virtual Worlds and Robotics
- 14. "Don't worry about turning your system over to the computer programmers. There won't be any in 1985, because the machine will be doing the job itself. The job will be automated, as intelligent computers program themselves." (1960) Herbert Simon one of the founding fathers of several of today's important scientific domains, including artificial intelligence, information processing, decision-making, problem-solving, attention economics, organization theory, complex systems, and computer simulation of scientific discovery. Industrialization of Software Engineering, Internet services and Enterprise IT (1996 onwards) “…The increasing complexity of programming work associated with a new and more powerful generation of computers had overwhelmed the technical and managerial ability of software groups. Software was late, over budget, lacked features, worked inefficiently, and was unreliable. Something to be called “software engineering” was proposed as the solution to the crisis. …” – as concluded by NATO Conference on Software Engineering, held in Garmisch, Germany in 1968. Thomas Haigh - “Crisis, What Crisis?” Reconsidering the Software Crisis of the 1960s and the Origins of Software Engineering (2010)
- 15. Economics of mass production embrace the human brainchild enabled by computers: “Software” Difference Area Hardware Software Human Factors Major Life-cycle Cost Source Development, manufacturing Maintenance and Evolution Training and operations labor Ease of Changes Generally difficult Good within architectural framework Very good, but people-dependent Change Process Manual, labor-intensive, expensive Electronic, inexpensive Needs retraining, can be expensive User-tailorability/Friendliness Generally difficult, limited options Technically easy; mission-driven Technically easy; mission-driven Sub-setability/Divisibility Inflexible lower limit Flexible lower limit Smaller increments easier to introduce Underlying Science Physics, chemistry, continuous mathematics Discrete mathematics, linguistics/programming Behavioral sciences Testing By test organization; much analytic continuity By test organization; little analytic continuity Directly by users Adapted from
- 16. 1975 - Software Engineering of IBM OS/360 defies industrial (large scale manufacturing) management: increasing the team size also increases the time to complete the work Culprit: Communication of thinking among software engineers combined with lack of job specific tools Domain specific Language(s) Domain(s) specific Models(s) Multiple perspectives Multiple tools ~ Industrialization of Software Years to Develop Software, Hardware HW Thousands of source lines of code (KSLOC) SW Today: Massive software projects become possible, with exponentially growing delivery time. Culprit: the cone of requirements uncertainty ( inability to imagine the end product amplified by the easiness to change the current one) "adding manpower to a late software project makes it later". Fred Brooks Stafford Beer, “The disregarded tools of modern man”, lectures, 1973
- 18. “Mapping the value of employee collaboration” McKinsey Quarterly 2006, Number 3
- 20. Context Factor Attribute of Relevant theoretical concept Im- pact * Explanation Technology Hardware cost Product Relative advantage + Lintel runs on commodity hardware Software cost Innovation Relative advantage + OSS operating systems are ??free?? Reliability Product Relative advantage +/- Varying perceptions of OSS platform reliability Availability of 3rd party apps Product Compatibility Network effects + Prerequisite to adoption, depends on platform popularity Portability of own apps Product Compatibility Switching costs +/0 Increases adoption where such apps exist Skills of existing IT workers Product Compatibility Switching costs +/- Increases adoption if and only if existing skills are compatible Fit to task Product Compatibility +/0 Increases adoption for certain tasks Difficulty in administra-tion Product Complexity - Perceived complexity decreases adoption Ease of ex-perimenting Innovation Trialability + Reduces risk Organization IT capital budget Innovation Slack - Large budgets alllow more choice of expensive options IT staff time Innovation Slack + Slack required to evaluate new technologies Innovativeness of IT organization Innovation Innovativeness + More innovative firms take more risks, want to be ??cutting edge?? Worker experience with new platform Product Boundary spanning + Linux knowledge that workers bring to organization prior to adoption Environment Industry maturity Innovation Industry life cycle - Infant industries not committed to old ways Availability of skilled IT workers Product Support infrastructure Network effects + Availability essential to adoption, more likely with popular platforms Availability of external support services Innovation Support infrastructure Sponsorship + Support needed to run in critical environments and to reassure management Platform long-term viability Product "Angry orphan" (switching costs) + Organizations avoid (re)investment in technologies that may become unsupported WHY FIRMS ADOPT OPEN SOURCE PLATFORMS: A GROUNDED THEORY OF INNOVATION AND STANDARDS ADOPTION (HBS and MIT Sloan, 2003)
- 21. Fluid Intelligence (perception and reasoning) KnowledgeExperience •Knowledge is a fluid mix of framed experience, values, contextual information and expert insight that provides a framework for evaluating and incorporating new experiences and information. •In Enterprises it often becomes embedded not only in documents or repositories but also in organizational routines, processes, practices and norms. Crystallized Intelligence (educational & cultural interactions) Skills Tacit knowledge drives learning Procedural knowledge stresses executive functions
- 22. John B. Carroll's three stratum model of cognitive abilities. -fluid intelligence (Gf), -crystallized intelligence (Gc), - general memory and learning (Gy), - broad visual perception (Gv), - broad auditory perception (Gu) - broad retrieval ability (Gr), - broad cognitive speediness (Gs) - processing speed (Gt). Tacit knowledge Fluid Intelligence (inductive & deductive reasoning) KnowledgeExperienceSkills Crystallized Intelligence (educational & cultural interactions) Internal representation of knowledge resulting from the slow process of learning Cognitive Neuroscience (from Prof. Garzzaniga) Cognitive Computing DL simulates Gy, Gv & Gu while inherently apt at Gr, Gs &Gt
- 23. Qualitative Analysis Quantitative Analysis Hypothesis Formulation Factor/Dependencies Analysis Concepts Modeling Systems (dynamic, agents) Modeling & Simulation Consistent/coherent visual design Argumentation support Information diffusion Research overwhelmingly shows that human have preferences and poor objectivity 95 % time, people fail to alter their behavior through conscious effort alone High motivation it simply takes too much mental energy and vigilance Computers are inherently logical and controlled They can channel the creativity of the “Fast brain” through visuals While engaging the “Slow brain” to think broad and deep They retain this work as an external, automated, and objective mind They can show it back in different ways, helping to iterate Economics … Enaction …
- 24. Summary Industrialization of Software Engineering, Internet services and Enterprise IT (1996 onwards) Cognitive computing for collaborative creative work in telepresence (analog or digital – does it matter?) Lucia.Gradinariu@lggsolutions.com - Synchronous collaboration on creative work (conceptual - The reasoning software engineer - The memory – consumer-entertainment product, executive functions impairments